Goliath Super Intelligence
InternationalOctober 19, 20262 min read

Big Tech AI Assistants Face Language and Infrastructure Hurdles in the Global South

Meta’s Muse and OpenAI’s Dots are expanding worldwide, but limited internet access, multilingual demands and costly services could widen digital gaps in developing economies.

Meta’s Muse and OpenAI’s Dots are marketed as personal assistants that can reserve tables, order groceries and handle other routine tasks. App analytics firm AppMagic reported that Muse has been downloaded nine million times in the United States and Canada since its September 8 launch, with about 1.8 million daily active users. Meta’s chief AI officer, Alexandr Wang, said the company is eager to bring personal agents to billions of people worldwide, according to the Financial Times.

South African startup ManyHands has already built a WhatsApp-based assistant called Manny that can place grocery orders, schedule appointments and process payments. Founder Anton Moulder explained to Rest of World that training the assistant to converse fluently in South Africa’s eleven official languages has proven difficult, highlighting the linguistic complexity that larger models must overcome to serve diverse markets.

Researchers note that large language models still lag behind in many non-English languages, limiting the benefits for speakers of those tongues. Payal Arora, a professor at Utrecht University and founder of the Inclusive AI Lab, told Rest of World that linguistic inclusion is fundamentally a matter of power. She added that agents need to grasp local cultural cues, family dynamics and informal networks to avoid imposing Western assumptions, a point echoed by independent policy researcher Usha Raman.

Network capacity could become a bottleneck as agents generate far more data traffic than human users. Cisco describes agents as “network power users,” producing up to 450 % more traffic per task in its tests, while Cloudflare observed a 1,700 % rise in daily AI-agent requests over the past year. University of Michigan professor Kentaro Toyama warned that poor connectivity in many developing regions would hamper both user interactions and the agents’ ability to complete online transactions.

Internet penetration remains low in many parts of Africa, with only 36 % of the continent online in 2025 compared with over 90 % in Europe, according to the International Telecommunication Union. Jane Munga of the Carnegie Endowment warned that AI assistants could deepen existing inequalities, especially when services like OpenAI’s Dots cost at least $100 per month, a price many users in the Global South cannot afford. Even free agents may be out of reach for those without reliable broadband.

Local merchants would need to adapt their websites for bot compatibility, a requirement that small, offline vendors often lack the resources to meet. Toyama noted that agents could increase purchase volumes and generate more return requests, putting additional strain on customer-service operations. Mercy King’ori, policy manager for Africa at the Future of Privacy Forum, told Rest of World that when agents are hosted abroad, national regulators may struggle to oversee data handling and consumer protection.

Sources

  1. Will developing nations be able to handle millions of AI agents? Rest of World

More reports

Industry · October 19, 2026 · 2 min

Anthropic embeds nonprofit METR as internal AI watchdog with publishing rights

The METR team can access Claude’s hardware and systems, publish findings without Anthropic’s veto, and operate under redaction rules while the company readies a multitrillion-dollar IPO.

Industry · October 19, 2026 · 2 min

Disney’s Reverse Acqui-Hire of Character.AI Uses Bankruptcy to Acquire Talent and Technology

Through a two-stage deal involving a $5.5 million payment, a refundable deposit and a possible $12.5 million license fee, Disney seeks Character.AI’s code while the bankrupt entity retains its lawsuits and state probes.

International · October 19, 2026 · 2 min

Anthropic earmarks five percent of equity for public-benefit fund ahead of IPO

The AI startup plans to allocate shares worth roughly $48 billion to a fund for large charitable donations, with additional contributions tied to future valuation gains.